https://github.com/context-hub/generator
ctx: The missing link between your codebase and your LLM. Context as Code (CaC) tool with MCP server inside.
https://github.com/context-hub/generator
context instructions llm mcp mcp-server php prompt-engineering vibe-coding
Last synced: 28 days ago
JSON representation
ctx: The missing link between your codebase and your LLM. Context as Code (CaC) tool with MCP server inside.
- Host: GitHub
- URL: https://github.com/context-hub/generator
- Owner: context-hub
- License: mit
- Created: 2025-03-08T08:37:28.000Z (2 months ago)
- Default Branch: main
- Last Pushed: 2025-04-10T18:46:42.000Z (28 days ago)
- Last Synced: 2025-04-11T00:44:17.683Z (28 days ago)
- Topics: context, instructions, llm, mcp, mcp-server, php, prompt-engineering, vibe-coding
- Language: PHP
- Homepage: https://docs.ctxgithub.com
- Size: 1.07 MB
- Stars: 54
- Watchers: 5
- Forks: 5
- Open Issues: 27
-
Metadata Files:
- Readme: README.md
- Funding: .github/FUNDING.yml
- License: LICENSE
Awesome Lists containing this project
- awesome-AI-driven-development - ctx - a tool made to solve a big problem when chatting with LLMs like ChatGPT or Claude: giving them enough context about your project. (Uncategorized / Uncategorized)
README
# ctx: The missing link between your codebase and your LLM. Context as Code (CaC) tool with MCP server inside.
**ctx** is a tool made to solve a big problem when chatting with LLMs like ChatGPT or Claude: **giving them enough context about your project**.
> There is an article about Context Generator
> on [Medium](https://medium.com/@butschster/context-not-prompts-2-0-the-evolution-9c4a84214784) that explains the
> motivation behind the project and the problem it solves.
Instead of manually copying or explaining your entire codebase each time, ctx automatically builds neat, organized context files from:
- Code files,
- GitHub repositories,
- Git commits and diffs
- Web pages (URLs) with CSS selectors,
- and plain text.
It was created to solve a common problem: **efficiently providing AI language models like ChatGPT, Claude with necessary
context about your codebase.**
## Why you need this
When you're using AI in development, contextt isn't just helpful — it's everything.
- **Code Refactoring Help**: Need AI assistance refactoring messy code? **ctx** creates clean, structured documents with all necessary code files.
- **Multiple Iteration Development**: Working through several iterations with an AI helper requires constantly updating
the context. **ctx** automates this process.
- **Documentation Generation**: Transform your codebase into comprehensive documentation by combining source code with
custom explanations. Use AI to generate user guides, API references, or developer documentation based on your actual
code.
- **Seamless AI Integration**: hanks to built-in MCP support, you can [connect](https://docs.ctxgithub.com/mcp-server.html) Claude AI directly to your codebase, allowing
for real-time, context-aware assistance without manual context sharing.
## How it works
1. Gathers code from files, directories, GitHub repositories, web pages, or plain text.
2. Targets specific files through pattern matching, content search, size, or date filters
3. Applies optional modifiers (like extracting PHP signatures without implementation details)
4. Organizes content into well-structured markdown documents
5. Saves context files ready to be shared with LLMs
6. Optionally serves context through an MCP server, allowing AI assistants like Claude to directly access project
information
**With ctx, your AI conversations just got a whole lot smarter—and easier.**
# Quick Start
Getting started with Context Generator is straightforward. Follow these simple steps to create your first context file
for LLMs.
## 1. Install Context Generator
Download and install the tool using our installation script:
```bash
curl -sSL https://raw.githubusercontent.com/context-hub/generator/main/download-latest.sh | sh
```
This installs the `ctx` command to your system (typically in `/usr/local/bin`).
> **Want more options?** See the complete [Installation Guide](https://docs.ctxgithub.com/getting-started.html) for alternative installation methods.
## 2. Initialize a Configuration File
Create a new configuration file in your project directory:
```bash
ctx init
```
This generates a `context.yaml` file with a basic structure to get you started.
> **Pro tip:** Run `ctx init --type=json` if you prefer JSON configuration format.
> Check the [Command Reference](https://docs.ctxgithub.com/getting-started/command-reference.html) for all available commands and options.
## 3. Describe Your Project Structure
Edit the generated `context.yaml` file to specify what code or content you want to include. For example:
```yaml
documents:
- description: "User Authentication System"
outputPath: "auth-context.md"
sources:
- type: file
description: "Authentication Controllers"
sourcePaths:
- src/Auth
filePattern: "*.php"
- type: file
description: "Authentication Models"
sourcePaths:
- src/Models
filePattern: "*User*.php"
```
This configuration will gather all PHP files from the `src/Auth` directory and any PHP files containing "User" in their
name from the `src/Models` directory.
#### Need more advanced configuration?
- Learn about [Document Structure](https://docs.ctxgithub.com/documents.html) and properties
- Explore different source types like [GitHub](https://docs.ctxgithub.com/sources/github-source.html), [Git Diff](https://docs.ctxgithub.com/sources/git-diff-source.html),
or [URL](https://docs.ctxgithub.com/sources/url-source.html)
- Apply [Modifiers](https://docs.ctxgithub.com/modifiers.html) to transform your content (like extracting PHP signatures)
- Discover how to use [Environment Variables](https://docs.ctxgithub.com/environment-variables.html) in your config
- Use [IDE Integration](https://docs.ctxgithub.com/getting-started/ide-integration.html) for autocompletion and validation
## 4. Build the Context
Generate your context file by running:
```bash
ctx
```
The tool will process your configuration and create the specified output file (`auth-context.md` in our example).
> **Tip**: Configure [Logging](https://docs.ctxgithub.com/advanced/logging.html) with `-v`, `-vv`, or `-vvv` for detailed output
## 5. Share with an LLM
Upload or paste the generated context file to your favorite LLM (like ChatGPT or Claude). Now you can ask specific
questions about your codebase, and the LLM will have the necessary context to provide accurate assistance.
Example prompt:
> I've shared my authentication system code with you. Can you help me identify potential security vulnerabilities in the
> user registration process?
> **Next steps:** Check out [Development with Context Generator](https://docs.ctxgithub.com/advanced/development-process.html) for best practices on
> integrating context generation into your AI-powered development workflow.
That's it! You're now ready to leverage LLMs with proper context about your codebase.
## 6. Connect to Claude AI (Optional)
For a more seamless experience, you can connect Context Generator directly to Claude AI using the MCP server:
There is a built-in MCP server that allows you to connect Claude AI directly to your codebase.
Point the MCP client to the Context Generator server:
```json
{
"mcpServers": {
"ctx": {
"command": "ctx server -c /path/to/your/project"
}
}
}
```
> **Note:** Read more about [MCP Server](https://docs.ctxgithub.com/mcp-server.html) for detailed setup instructions.
Now you can ask Claude questions about your codebase without manually uploading context files!
## JSON Schema
For better editing experience, Context Generator provides a JSON schema for autocompletion and validation in your IDE:
```bash
# Show schema URL
ctx schema
# Download schema to current directory
ctx schema --download
```
> **Learn more:** See [IDE Integration](https://docs.ctxgithub.com/getting-started/ide-integration.html) for detailed setup instructions for VSCode,
> PhpStorm, and other editors.
## Full Documentation
For complete documentation, including all available features and configuration options, please visit:
https://docs.ctxgithub.com
---
### License
This project is licensed under the MIT License.